1
1

btl/ugni: do not return a frag from sendi if an endpoint is waitlisted

This fixes a hang that can occur when running bandwidth tests.

Signed-off-by: Nathan Hjelm <hjelmn@lanl.gov>
Этот коммит содержится в:
Nathan Hjelm 2017-03-14 10:14:13 -06:00
родитель 2e42b0afbd
Коммит 6b210fa2c4

Просмотреть файл

@ -118,6 +118,13 @@ int mca_btl_ugni_sendi (struct mca_btl_base_module_t *btl,
size_t packed_size = payload_size; size_t packed_size = payload_size;
int rc; int rc;
if (OPAL_UNLIKELY(opal_list_get_size (&endpoint->frag_wait_list))) {
if (NULL != descriptor) {
*descriptor = NULL;
}
return OPAL_ERR_OUT_OF_RESOURCE;
}
do { do {
BTL_VERBOSE(("btl/ugni isend sending fragment from %d -> %d. length = %" PRIu64 BTL_VERBOSE(("btl/ugni isend sending fragment from %d -> %d. length = %" PRIu64
" endoint state %d", OPAL_PROC_MY_NAME.vpid, endpoint->peer_proc->proc_name.vpid, " endoint state %d", OPAL_PROC_MY_NAME.vpid, endpoint->peer_proc->proc_name.vpid,
@ -134,8 +141,7 @@ int mca_btl_ugni_sendi (struct mca_btl_base_module_t *btl,
} }
assert (packed_size == payload_size); assert (packed_size == payload_size);
if (OPAL_UNLIKELY(NULL == frag || OPAL_SUCCESS != mca_btl_ugni_check_endpoint_state (endpoint) || if (OPAL_UNLIKELY(NULL == frag || OPAL_SUCCESS != mca_btl_ugni_check_endpoint_state (endpoint))) {
opal_list_get_size (&endpoint->frag_wait_list))) {
break; break;
} }